The Silicon Sensors for the High Granularity Calorimeter of CMS
Total Page:16
File Type:pdf, Size:1020Kb
The Silicon Sensors for the High Granularity Calorimeter of CMS Peter Paulitsch∗, on behalf of the CMS Collaboration Austrian Academy of Sciences, Institute of High Energy Physics (HEPHY), Nikolsdorfer Gasse 18, 1050 Wien, Austria Abstract The installation of the High-Luminosity Large Hadron Collider (HL-LHC) presents unprecedented challenges to experiments like the Compact Muon Solenoid (CMS) in terms of event rate, integrated luminosity and therefore radiation exposures. To cope with this new environment, new detectors will be installed during the CMS Phase 2 Upgrade, including the replacement of the calorimeter endcaps with the "High Granularity Calorimeter" (HGCAL), which contains silicon sensors and scintillators as active elements. The silicon sensors will be produced in an 8" wafer process, which is new for high-energy physics, so it demands extensive quality verification. A first batch of prototype sensors underwent electrical tests at the institutes of the CMS Collaboration. Testing revealed major problems with the mechanical stability of the thin backside protective layer, that were not seen in earlier 6" prototypes produced by a different backside processing method. Following these results, the HGCAL group introduced the concept of "frontside biasing", allowing testing of the sensors without exposing its backside, verified the applicability, and adapted the prototype design to apply this method in series production. Keywords: Compact Muon Solenoid, Large Hadron Collider, High-Luminosity, High Granularity Calorimeter, large area, silicon pad sensors 1. Introduction lowing for efficient mitigation of pileup and particle- flow calorimetry. During the Phase-2 Upgrade (2025 to 2027), the Large Hadron Collider (LHC) will be upgraded to the High-Luminosity LHC (HL-LHC) [1]. The HL-LHC will have a factor 5–7 higher instantaneous luminosity hadronic calorimeter compared to the end of LHC operation, resulting in a proportionally higher pileup and a factor 10 increase in integrated luminosity (3000 fb−1) over 10 years of oper- ation. As a result, unprecedented levels of radiation and particle shower densities will affect experiments such as the Compact Muon Solenoid (CMS) [2]. At these high collision rates, the overlap of particle showers will not be negligible any more, so detectors with increased spatial resolution are needed to distinct different show- ers. To address these challenges, the CMS Collabora- electromagnetic tion will upgrade its subdetectors including a replace- calorimeter ment for the existing endcap calorimeters with the new Figure 1: Location of the HGCAL at the CMS endcaps [4]. High Granularity Calorimeter (HGCAL) [3], as shown arXiv:2002.11449v3 [physics.ins-det] 29 Jun 2020 in Figures1 and2. The calorimeter will utilize about The HGCAL will be a sandwich calorimeter and 1 2 30000 sensor modules covering more than 620 m , al- will include an electromagnetic part (Calorimeter Endcap - Electromagnetic, CE-E) and a hadronic part ∗ Corresponding author (Calorimeter Endcap - Hadronic, CE-H). While the Email address: [email protected] (Peter Paulitsch) 1This value is preliminary, it may be changed for optimized active sensing elements of the electromagnetic part calorimeter coverage by partial sensors, see Chapter 3.3 will be entirely made of silicon sensors, the hadronic Preprint submitted to NIM A July 1, 2020 elements will implement silicon just for the inner high increased radiation levels to cope with higher radiation radiation domain. At the outer regions with lower damage induced leakage currents (Table1). radiation levels, plastic scintillators coupled to silicon photomultipliers [3] will be used, as shown in Figure2. Table 1: Active thicknesses dact, number of channels, maximum ex- pected fluences Φneq (normalized to 1 MeV neutron equivalent), and This article focuses on the silicon parts of the CE-E and −1 maximum expected total ionizing dose (TID) at 3000 fb [3]. Φneq CE-H sections. is defined as the number of neutrons passing per sensor area and typ- ically describes lattice displacement damage. The energy spectrum is normalized to 1 MeV monoenergetic neutrons. The TID is the dose by charged particles, and typically describes damages in oxide layers, caused by generating and accumulating immobile charges. full-size -2 dact (µm) channels Φneq (cm ) TID (Gy) 120 432 (HD) 7:0 × 1015 1 × 106 200 192 (LD) 2:5 × 1015 2 × 105 300 192 (LD) 5:0 × 1014 3 × 104 Hamamatsu will manufacture the 120 µm sensors in an epitaxial process, whereas the thicker will be pro- duced in a float-zone process, as shown in Figure3. The shape of a full sensor is hexagonal (see Figure4) because a hexagon is the largest seamlessly tileable, regular shape on a circular wafer. This maximizes the wafer-area usage, reduces the number of necessary sen- sor tiles, hence decreases costs. 320/FZ290 6-inch 200/FZ200 8-inch active thin backside implant 200µm 200µm physical active 290µm 320µm physical thick backside implant 300/FZ300 8-inch 300/epi120 8-inch active 120µm active thick backside implant 300µm 300µm 300µm physical Figure 2: Schematic cross section of an endcap sector [5]. physical thin backside implant Figure 3: Sensor cross sections at different manufacturing processes. "epi" denotes an epitaxial process, "FZ" stands for "float-zone". For comparison, one 6" processed sensor cross section is shown. The thin 2. Silicon sensors for the HGCAL backside implant of the FZ200 and FZ300 processes is about 1 µm thick. The silicon sensors of the HGCAL will be produced in an 8" process [3], in contrast to earlier applications Consequently, the active sensing elements also have in high-energy physics, which used 6" processes. The a hexagonal shape, except for some irregular cells at the leap towards the 8" process decisively reduces produc- sensor edges and corners. Studies from RD502 [6] have tion costs and sensor testing efforts. However, this pro- shown that n-in-p materials have better charge collec- cess is new to large-area sensors for high-energy physics tion than p-in-n, so like in the CMS Tracker [7], n-in- and therefore brings new challenges in terms of radi- p diodes were chosen also for the HGCAL. For the ation hardness, high-voltage stability, and other issues 200 µm and 300 µm thick sensors, the so-called "Low- like sensor backside sensitivity (see Chapter 3.2). Three Density" (LD) design with 192 full-size channels is cho- different sensor thicknesses will be deployed: 120 µm, sen. To keep the diode (cell) capacitances and therefor 200 µm and 300 µm. Radiation damage increases the sensor leakage current. Thinner sensors draw less leak- 2Radiation hard semiconductor devices for very high luminosity age current, so these will be utilized in regions with colliders, http://rd50.web.cern.ch/rd50/ 2 series noise in the preamplifier low, the 120 µm thick carrier which holds a set of probe tips in a defined geo- sensors have smaller cells and therefore 432 full-size metric layout. channels, the "High-Density" (HD) design as shown in Characterization of the sensors is primarily fulfilled by Figure4. To control the electric fields and to reduce the measuring current versus voltage (IV) and capacitance versus voltage (CV) curves [10][11]. For all individual sensor cells on the same wafer, these measurements al- low extracting parameters like the full depletion voltage (Vfd), full depletion capacitance (Cfd), or the breakdown voltage (Vbd). 3.2. Sensor backside sensitivity and scratch tests Diode current characteristics on 8" HGCAL sensor prototypes showed a degradation in terms of an increas- ing number of cells with early breakdowns as shown in Figure5. This degradation only occured after sen- sor handling procedures were done. Repeated mea- surements without handling procedures in between re- mained stable. Because earlier 6" prototypes did not show this behavior, we suspected that the thin back- side metalization of the 8" prototypes might cause early breakdowns and increased sensor currents. The previ- ous 6" prototypes had a backside with a thick ("deep- diffused") field stop implant, seen on Figure3. The purpose of the highly doped field stop is to have zero electrical field at the beginning of the backside metal- Figure 4: 432-channel (HD) full wafer design, thickness 120 µm. The ization. This stops the spread of the field also at high red squares at the wafer periphery mark the four test structure half bias voltage. moons [8]. leakage current at the edge region, the sensor features two n++ doped guard rings; the inner one is connected to ground and the outer one is floating. A p++ doped edge ring is placed at the sensor edge, as shown in Fig- ure 11. This edge ring brings the edge region to the same potential as the p++ doped backside implant. Oth- erwise, microscopic cracks at the cutting edges would lead to field spikes which may cause breakdowns. 3. Sensor testing 3.1. Test system Figure 5: Sensor degradation after repeated handling steps, 192- channels, 200 µm active thickness. New breakdown cells (marked by During the production phase, multiple CMS institutes circles) appeared after each handling step. Cells with values of "0.00" will do quality control of 1–2 % of all produced sen- have bad contacts. The first test ("From HPK") was done by Hama- sors. Because of the large quantities of sensors pro- matsu before shipment. duced, CMS needs fast and widely automatized test equipment. Probe-cards are a common standard in the To investigate these problems with the fragile back- semiconductor industry for large-scale quality control side, scratch tests with a tungsten carbide needle were because they allow fast and easily reproducible mea- performed on test structures [12][13]. Using a needle surements. Therefore the collaboration developed an manipulator, the needle tip (50 µm diameter) was kept open-source full wafer probe-card and switching sys- hovering over the test structure diode, which was fixed tem, called "ARRAY" [9].